app.py

# app.py

import os
import gradio as gr
from llama_index.readers.file import PDFReader
from llama_index.core import VectorStoreIndex
from llama_index.llms.openai import OpenAI
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.core.chat_engine import ContextChatEngine
from llama_index.core.llms import ChatMessage, MessageRole

# --- Use HF secret for OpenAI API key ---
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
if not OPENAI_API_KEY:
    raise EnvironmentError("OPENAI_API_KEY is not set in environment variables.")

os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY

# --- Global Chat Engine ---
chat_engine = None
resume_text = ""

# --- Follow-up Generator ---
def generate_followups(user_question, resume_excerpt):
    llm = OpenAI(model="gpt-3.5-turbo")
    prompt = (
        f"The user asked: '{user_question}'\n\n"
        f"Based on their resume content:\n{resume_excerpt}\n\n"
        "Suggest 1–2 intelligent follow-up questions they might ask next. "
        "Format it like:\n\n🔁 Follow-up Suggestions:\n- ...\n- ..."
    )
    try:
        response = llm.complete(prompt)
        return response.text.strip()
    except Exception:
        return "🔁 Follow-up Suggestions:\n- Unable to generate follow-ups at this time."

# --- Upload & Index Resume ---
def upload_and_index(file):
    global chat_engine, resume_text
    try:
        reader = PDFReader()
        documents = reader.load_data(file=file.name)
        resume_text = "\n".join(doc.text for doc in documents[:1])

        index = VectorStoreIndex.from_documents(documents)
        retriever = index.as_retriever(similarity_top_k=3)
        memory = ChatMemoryBuffer.from_defaults(token_limit=3000)
        llm = OpenAI(model="gpt-3.5-turbo")

        system_prompt = ChatMessage(
            role=MessageRole.SYSTEM,
            content=(
                "You are a professional resume assistant. Your goal is to help users understand their resume better and explore relevant opportunities. "
                "Use the uploaded resume and past questions to craft structured, formal answers. Include bullet points and use headings like Skills, Experience, or Education where appropriate."
            )
        )

        chat_engine = ContextChatEngine.from_defaults(
            retriever=retriever,
            memory=memory,
            llm=llm,
            system_prompt=system_prompt
        )

        return "✅ Resume indexed successfully. You can now chat with the assistant."

    except Exception as e:
        return f"❌ Error during indexing: {str(e)}"

# --- Chat Function ---
def chat_with_bot(message, history):
    if chat_engine is None:
        return "❌ Please upload and index your resume first."
    try:
        main_response = chat_engine.chat(message)
        followup = generate_followups(message, resume_text)
        return f"{str(main_response)}\n\n{followup}"
    except Exception as e:
        return f"❌ Error: {str(e)}"

# --- UI ---
with gr.Blocks() as iface:
    gr.Markdown("""
    <h1 style='text-align: center;'>Revue</h1>
    <p style='text-align: center; color: gray;'>Upload your resume and ask career-related questions. The assistant remembers context and suggests what to ask next.</p>
    <hr>
    """)

    with gr.Group():
        file_input = gr.File(label="📎 Upload Resume (PDF)", file_types=[".pdf"])
        upload_btn = gr.Button("📥 Upload & Index")
        status_box = gr.Textbox(label="Status", interactive=False)

    upload_btn.click(fn=upload_and_index, inputs=file_input, outputs=status_box)

    gr.Markdown("### 💬 Ask Questions About Your Resume")

    chatbot = gr.ChatInterface(
        fn=chat_with_bot,
        chatbot=gr.Chatbot(),
        textbox=gr.Textbox(placeholder="e.g. What are my key strengths for data science?", lines=1),
    )

iface.launch()